ATIM Model Predicts Operator Action Time in Nuclear Control Rooms with 3.17s Error

Human error in digital control rooms can have catastrophic consequences, yet predicting how long operators take to complete tasks has relied on costly experiments or opaque black-box models. Researchers from Tsinghua University and Chosun University have now introduced ATIM (ACT-R-Based Task Interface Model), a theory-driven framework that estimates operator action time directly from interface characteristics. The model decomposes total action time into four interpretable components—visual search, cognitive processing, motor execution, and interaction—integrating established principles from visual search theory, Fitts' law, and the ACT-R cognitive architecture. Interface features like target salience and semantic interference serve as direct inputs, eliminating the need for additional user studies.

To validate ATIM, the team used a dataset from a digital nuclear control room environment. They identified task types via data-driven clustering and calibrated separate parameter sets for novice and experienced users. On a held-out validation set, the model achieved a mean absolute error of just 3.17 seconds and a strong correlation with observed performance (r = 0.664). The results demonstrate that ATIM generalizes well to unseen data while maintaining full interpretability—a key advantage over deep learning approaches. This makes it a practical tool for ergonomic assessment and interface optimization in complex human-machine systems like nuclear plants, aviation cockpits, and emergency response centers.